Newest papers

Present but Hard to Read: Probing Latent Reasoning with Minimal Pairs

Aur Shalev Merin · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28498 · cited by 0 works (OpenAlex) · open-access PDF

Korean

움직이는 움직이는 숫자 추론은 종종 응답이 시작되는 모델의 상태를 읽고 값이 부재한다는 증거로 실패한 검사를합니다. 우리는 같은 서명 모델 Ouro-1.4B와 표준 변압기 Qwen 2.5-1.5B에서 움직이는 숫자로 질문에서 복사 된 숫자로 이것을 테스트합니다. 응답이 시작되는 숫자에서만 차이가있는 견적을 비교하면 마지막 견적 토큰에 존재한다는 것을 보여줍니다. 질문을 통해 공유 된 코드 (읽기 약 0.9)에서 움직이는 무작위 단어는 한 번에 한 번에 견적을 볼 수 있지만 일반적인 설정은 거의 읽지 않습니다 : 0.36 Ouro와 0.19 Qwen에서 5,249 훈련 질문 후에도. 숫자의 변화는 붐비

Machine translation (Meta M2M100); the English original follows.

Abstract

Probing studies of latent reasoning often read a model's state where the answer begins and take a failed probe as evidence that a value is absent. We test this with a digit copied from the question, in the looped model Ouro-1.4B and the standard transformer Qwen 2.5-1.5B. Comparing prompts that differ only in that digit shows it is present at the last prompt token, in a code shared across questions (readout about 0.9). A probe that sees one prompt at a time, the usual setting, barely reads it: 0.36 in Ouro and 0.19 in Qwen even after 5,249 training questions. The digit's change is small rather than crowded: random noise lands on equally busy directions, while a random word in its place moves the state far more and is readable. Since every varied token leaves a trace, a successful probe is ambiguous too. We give a protocol, including a synthetic null for generic mixing, for testing computed values against copies and mixtures of the inputs. In pre-registered studies of two-step problems it finds a small signal for the unwritten intermediate a+b just before Ouro writes the answer's last digit, and its sign follows the operation, though mixing alone produces much of that sign-following part. Holding that value at another problem's from the first loop on does not move the written digit toward the other problem's, while the same patch of the answer does: the value is readable but not used.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Security token, Computer science, Value (mathematics), Sign (mathematics), Arithmetic, State (computer science)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange